🔬 MXFP4 Quantization Visualizer

📊 Input BF16 Values

📏 Vec Size: 16

⚖️ Scale: 1.0

🎯 Output Comparison

BF16 Input (16-bit each)

0100000010100000

1011111110100000

0011111101100000

0100000010000000

0100000010100000

1011111110100000

0011111101100000

0100000010000000

0100000010100000

1011111110100000

0011111101100000

0100000010000000

0100000010100000

1011111110100000

0011111101100000

0100000010000000

MXFP4 Output (4-bit each)

0100

1010

0010

0110

0011

1001

0101

1100

0010

0110

1011

0101

1001

0011

1110

0110

💾 Original: 256 bits

📦 Compressed: 96 bits

🗜️ Ratio: 2.67:1

💰 Saved: 62.5%

🎲 MXFP4 E2M1 Lookup Table

Highlighted values show the quantized output mappings

⚙️ Quantization Process

Scale Factor: 1.0 (Max value: 4.0 → MXFP4 range: ±6.0)

📈 Quantization Error Analysis

MSE: 0.0938 | MAE: 0.250 | Max Error: 0.500

📊 Memory Analysis

📦

Original Size

64 bits

🗜️

MXFP4 Size

48 bits

💰

Memory Saved

16 bits